Testing models of legislative decision-making with measurement error: The robust predictive power of bargaining models over procedural models

Previous studies found that models emphasising legislative procedures make less accurate predictions of decision outcomes in the EU than the compromise model, a computationally simple variant of the Nash Bargaining Solution. In this journal, Slapin (2014) argues that this and other findings may be the result of measurement error. While acknowledging the importance of measurement error, we disagree with several assumptions in Slapin’s analysis, and show that his results are driven by an unrealistic assumption about how policy preferences are distributed among EU decision makers. We construct simulated data that more accurately reflect the distributions of policy preferences found in existing empirical evidence and suggested by theory, and demonstrate that measurement error is unlikely to have biased previous findings. If real-world decision-making took place according to the procedural model, then it would have made the most accurate predictions, even with data containing large amounts of measurement error. While this strengthens our confidence in previous studies’ findings, we explain why we should not discard procedural models.